Processing v Post Processing Issues?

[groberts]

I've been using DSS and Photoshop for a few years now and I'd like to think, have a reasonable understanding of the processes etc. but have always been bugged by the outcome when moving the image from the stacked DSS image to PS for post processing, which often does not seem to produce a better image despite my best efforts.  Since transitioning to the mono ZWO1600MM-Cool camera and especially narrowband the disparity seems even more clear.

A recent stacked Ha-image looked very promising in DSS after stacking - first image below.  But on moving the same stacked image to PS and carefully applying Levels and Curves the result - second image below - to my eye does not seem to be as good as that originally viewed in DSS.  

I'm wondering if I'm missing something here in the PS processing? Any advice on how to improve matters would be appreciated.

Thanks, Graham

 

 

        

[Alien 13]

I might be one of the few that thinks doing most if not all the adjustments in DSS gives far better results than using the saved file with no adjustments then imported to PS.

Alan

[steppenwolf]

The first thing I have noticed is that your background sky level is very low - I'm reading levels around 8 whereas a more realistic level is roughly between 22 and 25.

Not everyone agrees with me on this next point but do you adjust the levels in one swoop or in lots of small iterations? I find that lots of iterations with smaller adjustments at each stage gives better results.

 

[groberts]

Alan - the DSS version (which is a screen capture after stacking) is without any adjustments and, as I said, to me looks quite good as a starting point for post processing.  But when moved into PS goes dark and needs to be stretched - it has always been my understanding this is inevitable?

Steve, I tend to use three or four small stretching iterations + any number of small curve adjustments, trying to tease out specific detail on the relevant part of the curve, in this case the Pelican.  The version shown was done in this way with x3 linear stretches and x6 curves - I don't consider this to be the finished version but its intended to demonstrate my point.  Are you saying use even more, smaller steps will help?

Thanks, Graham            

[steppenwolf]

Hi Graham,

You seem to be doing the right thing with the iterations but perhaps a bit more aggression is required. Do you prefer this version or your base version:-

[groberts]

I knew there was a better image hiding in there, yours looks much better, specifically brighter and  more detail - just what I'm looking for.

What are you doing / what am I not doing / what do you mean by "aggression" specifically?  

Thanks, Graham

[gnomus]

The data is there so it must just be a matter of a more aggressive stretch.  When your done stretching, where does your histogram peak end up?  Mine usually ends up at round an eighth of a way across the histo - sometimes a bit more...

[groberts]

Yes that's what I've been thinking and getting a bit frustrated trying to find it!  

Being wary of clipping the blacks my histogram is closer to the left - you seem to be saying push the black pointer further right? I tend to take the black pointer to the base of the curve when stretching in Levels and the mid-point left to a value of about 1.30'ish a few times.  

I'm very encouraged by what Steve has uncovered in my data, just need to somehow get there myself now!

[steppenwolf]

In essence, Graham just keep on going and stretch the data a little more. My version was based in your first image and is the result of numerous Levels and Curves stretches. I adjusted the black point until I reached 24 using Levels in several iterations and then started working on the white point in several iterations, never exceeding a change of 10 points. I then used Curves to reduce the black point in four iterations then locked it (Ctrl Click on a suitable blackpoint) then adjusted the climb of the graph until I got a result that I liked. I ignored the black point for the final curves adjustment as the data was, after all, a copy of your posted image, not the original data.

[groberts]

Thanks Steve, I'll try a rework the image myself + experiment - seems like I'm being too timid?

The good news is that the data is there + if successful might need to revisit some old images and rediscover their true information.

Hope to get back with good news soon.

Regards, Graham

[steppenwolf]

Good luck, Graham, if I can help any further, just shout! I do agree though that the data is most certainly there and although I'd hesitate to use the word 'timid' (there is nothing worse than 'over-processing'!) but more 'aggression' (is it too naff to use the word 'boldness'?) will release the data, just be sure that you don't clip that black!

[ollypenrice]

Steve and I experimented with Ha data down here last month, comparing our approaches. As he knows, I'm not of the cautious camp with Ha!   Provided I'm dealing with a decent and deep stack I use a radical curve once the black and white points have been correctly set.

(BTW, this curve is quite widely used by some imagers. It isn't a piece of lunacy I invented, though I've invented a few of those!)

What it does is maximise the contrast throughout the nebulosity. The very steep and straight rising line introduces the highest possible separation between any two points on the original linear histogram (seen as the diagonal line here.)

The downside is that it gives larger stars, of course, but in NB this is less of an issue and, with Astrodon filters, hardly an issue at all.

The outgoing image is 'harder' and less natural looking than a log stretch performed in Levels but, I would argue, is more informative in that it separates shapes more effectively. In fact I like both approaches. I tend to go for the very high contrast approach because my Ha is usually destined to appear in the red channel where its contrasts will be heavily 'diluted' in the mix with red. Taking them into red with high contrast is, therefore, an advantage.

Olly

[Stub Mandrel]

17 hours ago, Alien 13 said:

I might be one of the few that thinks doing most if not all the adjustments in DSS gives far better results than using the saved file with no adjustments then imported to PS.

Alan

I agree.

I think this is because when you do adjustments in DSS you use ta full 32 bit depth, so the exported file has the maximum detail.

What I have found is it essential to make sure the saved file from DSS doesn't have any gross noise in the darker parts by making the stretch slightly less aggressive - this matters a lot if you have an LP gradient.

[ollypenrice]

I don't think 16/32 bit has much to do with it. The numbers sound good but look at the pictures. I think this is all down to the the shape of the curve.

Olly